Preferred Networks and SEMIFIVE to develop MN-Core L2000 AI inference chips

Nov 16, 2026
The successor to PFN's L1000 series, due in 2027, is planned to more than double memory bandwidth using stacked logic and DRAM.

Preferred Networks (PFN) and SEMIFIVE are collaborating to develop the MN-Core L2000 Series, PFN's next generation of chips designed for AI inference, the companies said in a press release. The series succeeds the MN-Core L1000 Series, which is due for release in 2027, and is planned to deliver more than twice its memory bandwidth.

PFN is developing the MN-Core L chips with a focus on memory bandwidth, which it sees as a key limit on generative AI inference performance, by vertically stacking logic chips based on its MN-Core architecture with DRAM. SEMIFIVE, which provides custom chip design, will handle design implementation and supply chain management using its 3D-IC design technology, and the logic chip will be manufactured on Samsung Foundry's platform.

"The MN-Core L2000 Series is the next generation of AI chips, designed to deliver even greater memory bandwidth than the MN-Core L1000 Series, taking another step toward overcoming bandwidth limitations, a major bottleneck for generative AI," said Daisuke Okanohara, PFN's co-founder and CEO.

We hope you enjoyed this article

Consider subscribing to one of our newsletters like Silicon Brief or Daily AI Brief.

Also, consider following us on social media:

Free newsletter

Silicon Brief

Weekly coverage of AI hardware developments including chips, GPUs, cloud platforms, and data center technology.

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more
Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.